Avalanche Treasury CEO warns AI agents may strain blockchain capacity

Blockchains have a space problem, and AI agents are about to make it worse. That’s the core argument from Bart Smith, CEO of Avalanche Treasury Company (AVAT), who used a recent podcast appearance to lay out why he thinks the next wave of autonomous software could fundamentally reshape which networks win institutional adoption.

Speaking on the July 9 episode of the Layer One podcast, Smith framed the issue in refreshingly blunt terms: blockchain capacity is finite, AI agents are multiplying fast, and networks that can’t isolate workloads will feel the squeeze first.

The AI agent bottleneck

By mid-February 2026, Avalanche had already recorded over 1,600 AI agents registered under the ERC-8004 standard, actively trading and executing on-chain tasks. Those agents don’t wait for gas prices to drop or congestion to clear. They transact continuously, and their volume compounds as more get deployed.

Smith’s point is that this dynamic makes the architectural differences between major chains genuinely consequential. Ethereum and Solana operate as high-traffic shared environments where every application competes for the same block space. When an AI-driven trading bot floods Ethereum with transactions during a volatile hour, every DeFi user on the network feels it through higher fees and slower confirmations.

Avalanche’s answer is its subnet model, which lets organizations spin up purpose-built Layer 1 blockchains that run independently. An AI-heavy trading subnet doesn’t cannibalize throughput on a tokenized real estate subnet. The workloads stay isolated.

AVAT’s institutional bet

Smith isn’t just an observer here. He’s running a company that went public on Nasdaq in June 2026 through a SPAC merger valued at roughly $675 million. AVAT has set an AVAX treasury target exceeding $1 billion, a figure that signals serious conviction in Avalanche’s long-term positioning.

The company’s strategy centers on building out infrastructure for enterprise clients. Avalanche currently supports approximately 80 live purpose-built Layer 1 blockchains, with projections to scale toward 200 custom institutional chains. Each one is tailored for a specific industry or use case, from real-world asset tokenization to compliance-heavy financial products.

For Smith, the AI agent explosion only accelerates this logic. Institutions already cautious about sharing infrastructure with unpredictable DeFi protocols will be even less enthusiastic about sharing it with thousands of autonomous agents they don’t control.

What this means for the competitive landscape

Smith’s argument targets a structural limitation rather than a performance metric. A congested Ethereum rollup still shares security and settlement with everything else on the network. Solana’s single-chain architecture means every transaction competes in the same validator queue.

For investors watching the Avalanche ecosystem, the key metric to track is how quickly those 80 live institutional L1s grow toward the 200 target. Each new deployment represents an enterprise client that has explicitly chosen isolation over shared infrastructure.

The broader signal is that AI agents are no longer a hypothetical variable in blockchain planning. They’re already on-chain, already consuming resources, and already forcing network architects to think differently about capacity.